Internet service is ubiquitous in today's network-connected world. Most users of the Internet secure their access through a subscription arrangement with an Internet Service Provider (ISP). ISPs facilitate communication between subscribers and the greater Internet by offering various modes of connection to their subscribers. The ISP forwards subscribers' outgoing messages to the Internet proper through a high-speed backbone connection. Likewise, subscribers' incoming messages are received from the high-speed backbone and are relayed to individual subscribers. By so doing the relatively high cost of access to the Internet is shared among a large number of subscribers, thus making Internet access affordable to individual users.
ISPs communicate with their subscribers through various kinds of connections, the vast majority of which require the use of some kind of modem. Examples include voice-band dial-up modems, satellite modems, wireless modems, cable modems and, in a growing number of instances, Digital Subscriber Line (DSL) modems. Given that newer modem technology is relatively young when compared, for example, to voice-band dial-up modem technology, the expense of keeping up to date with the latest developments in modem technology can constitute a significant cost burden for ISPs. One method that modem manufacturers have developed to help to reduce this cost burden is to use non-volatile memory programmable modems. Non-volatile memory programmable technology permits a subscriber's modem to be upgraded by remote control without the need for customer or even manual intervention. Special capabilities in these programmable modems permit the modem operating firmware to be upgraded by means of special signals that can be initiated by the ISP and transmitted over the communications line that connects the subscriber to the ISP's facilities. When a programmable modem receives such a signal, it suspends normal operation and enters a mode whereby certain program code is stored in the modem's non-volatile program memory. In some cases, this may be flash memory or Electrically Erasable Programmable Read Only Memory (EEPROM). Once code is stored in non-volatile program memory, it is semi-permanent in that the state of the memory is maintained even when modem power is turned off.
The state of affairs just described does not tell the whole story, however. It is unfortunate that communication between ISPs and subscribers' modems cannot be guaranteed to be 100% reliable. In fact, errors are a normal part of life for modems in their normal operating mode, and modem designers have developed many innovative techniques that lower the probability of such errors and that mitigate their effects when they do occur. In spite of these innovations, however, the probability of error during modem communication can never be zero. Additionally, many other sources of disruption in the communication process are outside the control of modem designers. Some of these sources of disruption are lightning strikes, damage to telephone lines caused by traffic crashes, storm damage to telephone lines, loss of power at the subscriber's premises, and the like. Therefore, while means exist for remotely reprogramming modems, the integrity of the process cannot be guaranteed, and means must be sought to cope with the result when the normal remote reprogramming process fails.
The problem is made especially difficult because of the chicken and egg nature of the whole process. Indeed, a subscriber's modem must have valid code on board in order to execute the instructions necessary to receive, process, and reload updated code being received from an ISP. However, given that non-volatile memory is relatively expensive it is normally necessary to erase all of the existing code stored in the memory before beginning the reprogramming process. During this interval, the firmware stored in the non-volatile memory is normally moved to random access memory (RAM). A microprocessor in the modem can then execute the firmware from the RAM image. If for some unhappy reason power to the modem is lost during the reprogramming process, the modem is left with no valid code with which to complete the process. Alternatively, if an error occurs in the sequence of many thousands of bits received by the subscriber's modem during reprogramming, the non-volatile memory will be programmed with invalid code, and the modem simply will not work.
The solution to this problem heretofore has been for the subscriber to return his modem to the manufacturer or dealer to have the unit reprogrammed. While manufacturers and ISPs may wish to believe that such situations occur only rarely, the loss of confidence both in the manufacturer and in the ISP on the part of the subscriber, not to mention the cost and general inconvenience to the subscriber, definitely are not good for business and should be avoided. What is needed is a way to recover from the catastrophic loss of firmware code stored in non-volatile memory that does not burden or inconvenience the subscriber.